Wednesday, November 20, 2013

Giving Hypothermia the Cold Shoulder

Modern resuscitation has many components – emphasis on pre-hospital CPR, distribution of defibrillators, cardiac catheterization ... and hypothermia.  And, now we're not so sure about that last one.

Therapeutic hypothermia is based on two randomized trials, published in 2002, enrolling a grand total of 352 patients.  Recommendations include rapid cooling after emergency stabilization to a target temperature between 32° and 34°C, and subsequent adoption of this practice spawned a cottage industry of expensive, specialized cooling devices.  But, these temperature targets were always somewhat arbitrary, and the question persisted as to whether lower temperatures in fact conferred a neurologic survival advantage.


Enrolling nearly a thousand patients, this randomized trial of out-of-hospital cardiac arrest randomized half to 33°C and half to 36°C.  Groups were well-balanced on typically expected features favoring survival – comorbidities, bystander CPR, arrest rhythm, and follow-up treatment.  And, after cooling for 24 hours, no difference was detected in early deaths, late deaths, or in any measure of cerebral or functional performance.  One impressive feature of this trial is the standardized withdrawal of care criteria, reducing patient heterogeneity through that mechanism.

None of this says "hypothermia doesn't work".  But, a sense of lacking in terms of a dose-response relationship between hypothermia and neurologic survival causes some concern.  Why did we observe such profound benefit in the early trials?  Erroneous rejection of the null hypothesis due to poor statistical power, unmeasured confounders, or something more sinister?  Adding in the pre-hospital hypothermia study from JAMA, both dose-dependent and time-dependent effects are now less certain.

The utility of hypothermia following ischemic insult seems biologically plausible, but, as the editorial comments, perhaps it's not so much hypothermia as the maintenance of normothermia.  Already part of modern post-stroke care, treating and preventing fever improves outcomes – it may simply be the observed benefits are due to intensive antipyresis, rather than hypothermia.

It seems still reasonable to use gentle cooling as a prophylaxis against hyperthermia, but more importantly, it is time, yet again, to reflect on how better evidence refines established practice.  Without continuing to recognize the limitations of our knowledge, we must caution ourselves against rushing to generalize implementation from small sample sizes (see: tPA in acute ischemic stroke).

"Targeted Temperature Management at 33°C versus 36°C after Cardiac Arrest"

"Temperature Management and Modern Post–Cardiac Arrest Care"


  1. Great post on what may be a practice changing article. Some questions / concerns regarding my interpretation of the paper.

    1) The two groups in the trial both underwent therapeutic hypothermia / temperature management, both were sedated, both had external cooling mechanisms to maintain temperature. Thirty six degrees is not by definition hypothermia, however, it is still slightly below physiologic requiring some external cooling. The study does not disprove the benefit of therapeutic hypothermia, there is no control that maintained only normothermia, it merely shows that the absolute cooling temperature is unclear.

    2) The study was powered to the primary outcome of death at 180 days - neurologic function was a secondary outcome and does not seem powered appropriately to state that there is no neurologic survival advantage.

    3) The study was powered to detect a relative risk reduction of 20%, could there still be a significant effect that, although the study was large, is still hidden by the lack of power

    4) Transitioning to real world application, I wonder if it will be harder to maintain such strict temperatures and, if at 36 degrees, there will be more chance of patients inadvertently slipping into hyperthermia, and, as such, have higher mortality.

    Just a few questions / really enjoy your blog - thanks for taking the time.


    Bill Soares, MD.
    Baystate Medical Center
    Springfield, MA

    1. All reasonable critiques – as with most work, it raises as many questions as it asks. Regarding lack of a true control group, it is true 36°C (+/- 1°C) is not normothermia. It is up to us to speculate whether there's a true physiologic difference given that single degree difference in temperature. Or, perhaps 35°C is the true temperature target … and both groups came down on either side of optimal cooling. Who knows!

      And, yes, there's still a chance a clinically significant effect were missed. Considering the advantages seen in the original trials, I don't think anyone will be rushing to completely throw out current practice without additional data. That being said, I would expect to start seeing folks shying away from the 32°C to 34°C range out of concern for other possible physiologic harms.

      Regardless, continued controversy keeps folks like me in business!


Comments on posts > 10 days old will be moderated; blame spammers.